# Gemini YOLO + Boring Maximum Utilization Guide
This guide explains how to combine Gemini CLI's YOLO mode with Boring MCP tools for maximum autonomous development efficiency(AUTO LOOP).
## Core Concept
```
Gemini YOLO = Engine (Execution Power)
Boring MCP = Toolbox (Professional Skills)
```
When combined, AI can automatically use Boring's 50+ tools to complete complex tasks.
---
## Prerequisites
### 1. Install Boring MCP
Ensure `~/.gemini/settings.json` contains:
```json
{
"mcpServers": {
"boring": {
"command": "uvx",
"args": ["boring-aicoding"]
}
}
}
```
### 2. Verify Setup
```bash
gemini "List all available boring tools"
```
---
## Common YOLO Commands
### 🔍 Code Search + Modification
```bash
gemini --yolo "Use boring_rag_search to find auth-related code, then fix the login bug"
```
### 🧪 Auto Test + Fix
```bash
gemini --yolo "Run boring_verify, if it fails fix the issues, repeat until passing"
```
### 📝 Code Review + Refactor
```bash
gemini --yolo "Run boring_code_review on all Python files in src/, then fix all suggestions"
```
### 🚀 Release Workflow
```bash
gemini --yolo "Execute /release-prep workflow, complete all steps"
```
---
## Advanced Combinations
### Full Auto Development Loop
```bash
gemini --yolo "
1. Use boring_rag_search to understand code structure
2. Read @fix_plan.md to find the next task
3. Implement that task
4. Use boring_verify to validate
5. If passing, mark [x] in @fix_plan.md
6. Repeat until all tasks complete
"
```
### Quality Gate Loop
```bash
gemini --yolo "
For each modified file:
1. boring_code_review audit
2. boring_perf_tips performance check
3. boring_vibe_check health score
Fix all issues until vibe score > 80
"
```
---
## Boring Tools Quick Reference
| Tool | Purpose | YOLO Usage Example |
|------|---------|-------------------|
| `boring_rag_search` | Semantic code search | Understand project structure |
| `boring_code_review` | AI code review | Auto-fix suggestions |
| `boring_vibe_check` | Project health score | Quality gate |
| `boring_verify` | Run verification | CI/CD automation |
| `boring_impact_check` | Impact analysis | Safe refactoring |
| `boring_security_scan` | Security scan | Prevent leaks |
| `boring_test_gen` | Auto generate tests | Increase coverage |
| `boring_doc_gen` | Auto generate docs | Keep docs synced |
---
## Automation Scripts
### Windows PowerShell
```powershell
# yolo_loop.ps1 - Automated YOLO loop
$MAX_LOOPS = 5
for ($i = 1; $i -le $MAX_LOOPS; $i++) {
Write-Host "=== Loop $i / $MAX_LOOPS ===" -ForegroundColor Cyan
# Execute one round
gemini --yolo "Complete the next [ ] task in @fix_plan.md, mark as [x] when done"
# Verify
boring verify
if ($LASTEXITCODE -eq 0) {
Write-Host "✅ Verification passed" -ForegroundColor Green
} else {
Write-Host "❌ Needs fixing" -ForegroundColor Red
}
Start-Sleep -Seconds 5
}
```
### Linux/Mac Bash
```bash
#!/bin/bash
# yolo_loop.sh - Automated YOLO loop
MAX_LOOPS=5
for i in $(seq 1 $MAX_LOOPS); do
echo "=== Loop $i / $MAX_LOOPS ==="
# Execute one round (with timeout to prevent infinite runs)
timeout 5m gemini --yolo "Complete the next task in @fix_plan.md"
# Verify
boring verify
if [ $? -eq 0 ]; then
echo "✅ Verification passed"
else
echo "❌ Needs fixing"
fi
sleep 5
done
```
---
## Common Scenario Commands
| Scenario | Recommended Command |
|----------|---------------------|
| Understand new project | `gemini --yolo "boring_rag_search analyze entire codebase"` |
| Fix bug | `gemini --yolo "Find and fix XXX issue"` |
| Refactor | `gemini --yolo "boring_impact_check + safe refactor"` |
| Release | `gemini --yolo "/release-prep"` |
| Write tests | `gemini --yolo "boring_test_gen generate tests for src/"` |
| Security check | `gemini --yolo "boring_security_scan and fix issues"` |
---
## Cautions
> ⚠️ **Warning**: YOLO mode skips all confirmations! AI can directly modify files and execute commands.
### Safety Recommendations
1. **Use Git**: Ensure all changes can be rolled back
2. **Set boundaries**: Clearly tell AI which files cannot be touched
3. **Execute in segments**: Break large tasks into smaller steps
4. **Regular reviews**: Manually review changes every few rounds
### Recommended Environments
- Docker containers
- Virtual machines
- Isolated test projects
- Projects with complete backups
